Statistical differentiation between direct and indirect effects of neuroleptics on negative symptoms

The differentiation between primary and secondary negative symptoms in schizophrenia (Carpenter et al. 1985) is an important issue. Path analysis allows to estimate statistically whether, and in which degree, the effect of a neuroleptic on negative symptoms is mediated by effects on positive, extrapyramidal, and depressive symptoms (Moller et al. 1995). If certain causal relationships are theoretically assumed - as proposed by Carpenter et al. (1985) - then path analysis can be applied to estimate the quantitative degree of these relationships, although the causal directions cannot be inferred from path analysis it self. Especially it can be estimated whether there is sufficient evidence for a ''direct effect'' of neuroleptic treatment on (primary) negative symptoms, an effect which is not mediated by positive, extrapyramidal, and/or depressive symptoms. We show the correspondence between the applied path model and several simple regression equations which can be estimated with standard statistical software. Moreover, we report some Monte Carlo studies showing that the results reported by Moller et al. (1995) - a ''direct effect'' of risperidone (6 mg) on negative symptoms compared with haloperidol (20 mg) - cannot be explained by a path model in which, everything else being equal, positive symptoms depend on negative symptoms instead of the other way around.